The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell operations for environment setup, session tracking, and project analysis. It executes local binaries and scripts located within the user's home directory (~/.claude/skills/gstack/bin/).
[DYNAMIC_EXECUTION]: The preamble uses eval and source to load environment variables and configuration states dynamically from the output of local scripts like gstack-slug and gstack-repo-mode.
[DATA_EXFILTRATION]: The skill includes functionality for usage telemetry and cross-machine session syncing via GBrain. Telemetry logs data to local files and remote endpoints, while GBrain can synchronize session artifacts to a private GitHub repository. These capabilities are gated by explicit user prompts and require opt-in consent before activation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from the local project and external repository metadata to populate its dashboard.
Ingestion points: Project files like VERSION and CLAUDE.md, as well as metadata retrieved via git and gh CLI tools (e.g., PR titles, branch names).
Boundary markers: External data is rendered directly into a console-based dashboard without the use of specific boundary delimiters.
Capability inventory: The skill has access to the Bash tool and can perform file system reads and remote repository queries via standard CLI utilities.
Sanitization: Content extracted from repository files and PR data is parsed using jq and cat for display without explicit sanitization or filtering of potential embedded instructions.

landing-report